In order to improve the precision of test date, a data fusion method determining the weighted value based on correlation function was presented. It was unnecessary for this algorithm to has any pre-defined knowledge about the measured data by these sensors, and it could adjust the fused sensor′s weight in time according to the variation of the data correlation degree. This method was flexible and simple with capability of dynamic adaptively and anti-jamming. The simulation shows that this method is effective and its fusion effects both in the precision and the fault tolerant surpass thet raditional weighting method of mean value. The application of the method to fault diagnosis of a rotary machine indicates that this method is feasible in the practical application.